A lot of bad ideas used to die because they were too expensive to execute.
I think that was an underrated form of quality control.
The old filter
If you wanted to build software, you needed engineers.
If you wanted a national campaign, you needed creative people, media buyers, production and money.
If you wanted a physical product, you needed prototypes, tooling, supply chains and inventory.
The cost of execution forced you to choose.
You could not pursue every idea at once, so some amount of judgment happened before the work started.
Founders called this focus. Investors called it capital allocation. A lot of the time it was simply scarcity doing the thinking for us.
AI weakens that filter
When research, writing, prototyping, design and routine software work get dramatically cheaper, more ideas survive long enough to become real things.
That is mostly good.
It also means the market gets flooded with things that were previously too weak to justify the cost of producing them.
We are already seeing the beginning of this in content. The marginal cost of another article, ad variation, image or email can approach zero. So people make more of them.
The obvious response is to celebrate the volume.
I think the harder question is what happens when volume stops being evidence of effort.
Effort was never quality, but it was a signal
A book that took three years to write could still be terrible. A product that consumed ten million dollars could still fail.
But effort and capital imposed friction.
Friction forced some decisions to happen before distribution.
When friction disappears, the judgment moves somewhere else.
The reader has to filter more. The customer has to filter more. Platforms have to filter more. Recommendation systems have to decide which of a million technically competent things deserves attention.
That makes trust, reputation and specificity more valuable.
The founder's job changes too
I used to think a founder's unfair advantage was the ability to get an unreasonable amount of work done.
There is still truth in that.
But if software can do more of the work, endurance becomes less differentiating.
The question shifts toward:
What should exist?
What is actually true?
Which customer is worth serving?
Which detail matters?
Where should the company refuse to automate?
What should we stop making?
Those are judgment questions.
The data already hints at the shift
Microsoft's 2026 Work Trend Index found that advanced AI users increasingly delegate routine execution while staying involved in direction, review and standards. It describes a progression from author to editor to director to orchestrator.
MICROSOFT WORK TREND INDEX · 2026 · AGENTS AND HUMAN AGENCY
I think that progression applies to founders too.
Not because founders stop making things.
Because the number of things they are capable of making grows faster than the number of things worth making.
Scarcity moved
Capital is still scarce. Attention is scarce. Great people are scarce. Distribution is scarce.
But one kind of scarcity is clearly moving: the ability to produce a first version of knowledge work.
That means the scarce asset becomes knowing which first version deserves a second.
Execution used to kill a lot of ideas before judgment had to.
We may miss that filter.
So we need to get better at replacing it deliberately.
